# Scene Graph Prediction with Limited Labels

## TL;DR

This paper presents a semi-supervised approach for scene graph prediction that effectively leverages limited labeled data and unlabeled images to improve relationship detection in visual knowledge bases.

## Contribution

It introduces a novel probabilistic label generation method using heuristics and a factor graph model, enabling training of scene graph models with minimal labeled data.

## Key findings

- Outperforms baseline methods by 5.16 recall@100 on PREDCLS
- Effective with as few as 10 labeled examples per relationship
- Relationship complexity correlates with method success (R^2=0.778)

## Abstract

Visual knowledge bases such as Visual Genome power numerous applications in computer vision, including visual question answering and captioning, but suffer from sparse, incomplete relationships. All scene graph models to date are limited to training on a small set of visual relationships that have thousands of training labels each. Hiring human annotators is expensive, and using textual knowledge base completion methods are incompatible with visual data. In this paper, we introduce a semi-supervised method that assigns probabilistic relationship labels to a large number of unlabeled images using few labeled examples. We analyze visual relationships to suggest two types of image-agnostic features that are used to generate noisy heuristics, whose outputs are aggregated using a factor graph-based generative model. With as few as 10 labeled examples per relationship, the generative model creates enough training data to train any existing state-of-the-art scene graph model. We demonstrate that our method outperforms all baseline approaches on scene graph prediction by 5.16 recall@100 for PREDCLS. In our limited label setting, we define a complexity metric for relationships that serves as an indicator (R^2 = 0.778) for conditions under which our method succeeds over transfer learning, the de-facto approach for training with limited labels.